Ffmpeg colour bugs

So.. Pretty nice guide you have here, but I think you slightly misunderstood *the situation*.
It is true that "all unspecified things the same" they stick to bt601, but the problem isn't really that there are assumptions about unknown still image formats (or better yet \*videos\*, because from png to jpg and webp I'm really skeptical any of them wouldn't have pretty rigorously documented default properties).
Everything is kind of a disaster because despite all of this, the program is then completely silent about such decisions (including within the final metadata) - and it is ***then*** the video players that have to try to guess what the color matrix is for yuv-rgb conversion.
And this is only one of the many shortcomings.

Long story short we have:
* files aren't [tagged](https://trac.ffmpeg.org/ticket/9167) with their colorspace information (and I'd like to remember the SD=601 and HD=709 convention otherwise)
* unless using [zscale](https://trac.ffmpeg.org/ticket/9693), forget about extracting videos to jpeg (unsure about the other way around?)
* output pngs have [wrong](https://trac.ffmpeg.org/ticket/10000) gamma (in one of the tags at least? depending on the software you may be good)
* colorspace and scale are also [busted](https://trac.ffmpeg.org/ticket/9772) in their own ways with linear gamma conversions

There are too what you could call "slight rounding errors" (the last three links [here](https://github.com/n00mkrad/flowframes/issues/157#issuecomment-1171774927)), but for once they seem pretty minor.

Commencez par consulter le guide d’encodage et les tickets FFmpeg liés 9167, 9693, 10000 et 9772, ainsi que la discussion Flowframes liée. L’issue couvre plusieurs questions liées à l’étiquetage des couleurs, à l’extraction, au gamma et à l’espace colorimétrique, mais ne nomme aucun fichier ni test du dépôt ; une contribution nécessiterait un périmètre défini et une définition de terminé convenue.

Rédigé par le modèle d'indexation à partir du texte de l'issue.